经典滤波算法去噪对比实验（Matlab实现）

# 一，经典滤波算法的基本原理 ###1，中值滤波和均值滤波的基本原理 参考以前转载的博客：http://blog.csdn.net/ebowtang/article/details/38960271 ###2，高斯平滑滤波基本原理 参考以前转载的博客：http://blog.csdn.net/ebowtang/article/details/38389747 # 二，噪声测试效果 ### 1，不同噪声效果 三幅图各噪声浓度分别是0.01 0.03,0.05（比如第一副图均是加入0.01的噪声浓度）    ### 2，实验代码 ~~~ %读入原始图像并显示 image_original=imread('dog.bmp'); figure(1) subplot(2,4,1); imshow(image_original); title('原输入图像'); axis square; %生成含高斯噪声图像并显示 pp=0.05; image_gaosi_noise=imnoise(image_original,'gaussian',0,pp); subplot(2,4,2); imshow(image_gaosi_noise); title('添加高斯噪声后图像'); axis square; %生成含椒盐噪声图像并显示 d=0.05; image_saltpepper_noise=imnoise(image_original,'salt & pepper',d); subplot(2,4,3); imshow(image_saltpepper_noise); title('添加椒盐噪声后图像'); axis square; %生成含乘性噪声图像并显示 var=0.05; image_speckle_noise=imnoise(image_original,'speckle',var); subplot(2,4,4); imshow(image_speckle_noise); title('添加乘性噪声后图像'); axis square; %原图像直方图 r=0:255; bb=image_original(:); pg=hist(bb,r); pgr1=pg/length(bb); subplot(245);bar(pgr1);title('源输入图像的直方图'); r=0:255; bl=image_gaosi_noise(:); pg=hist(bl,r); pgr2=pg/length(bl); subplot(246);bar(pgr2);title('高斯噪声污染后的直方图'); r=0:255; bh=image_saltpepper_noise(:); pu=hist(bh,r); pgr3=pu/length(bh); subplot(247);bar(pgr3);title('椒盐噪声污染后的直方图'); r=0:255; ba=image_speckle_noise(:); pa=hist(ba,r); pgr4=pa/length(ba); subplot(248);bar(pgr4);title('乘性噪声污染后直方图'); ~~~ # 三，椒盐噪声去除能力对比 ### 1，三大去噪效果 三幅图椒盐噪声浓度分别是0.01 0.03,0.05（比如第一副图均是加入0.01的椒盐噪声去噪对比）    ### 2，实现代码 ~~~

%读入原始图像并显示
image_original=imread('dog.bmp');
figure(1)
subplot(2,4,1);
imshow(image_original);
title('原输入图像');
axis square;

%生成含高斯噪声图像并显示
%pp=0.05;
%image_gaosi_noise=imnoise(image_original,'gaussian',0,pp);

%生成含椒盐噪声图像并显示
dd=0.05;
image_saltpepper_noise=imnoise(image_original,'salt & pepper',dd);

%生成含乘性噪声图像并显示
%var=0.05;
%image_speckle_noise=imnoise(image_original,'speckle',var);

image_saltpepper_noise_after1=medfilt2(image_saltpepper_noise,[3,3]);
subplot(2,4,2);
imshow(image_saltpepper_noise_after1);title('中值滤波去椒盐噪声效果图');
axis square;

h_gaosi1=fspecial('gaussian',3,1);
image_saltpepper_noise_after2=imfilter(image_saltpepper_noise,h_gaosi1);
subplot(2,4,3);
imshow(image_saltpepper_noise_after2);title('高斯平滑去椒盐噪声效果');
axis square;

image_saltpepper_noise_after3=wiener2(image_saltpepper_noise,[5 5]);
subplot(2,4,4);
imshow(image_saltpepper_noise_after3);title('维纳滤波去椒盐噪声效果');
axis square;

%原图像直方图
r=0:255;  
bb=image_original(:); 
pg=hist(bb,r);  
pgr1=pg/length(bb);  
subplot(245);bar(pgr1);title('源输入图像的直方图');

r=0:255;  
bl=image_saltpepper_noise_after1(:); 
pg=hist(bl,r);  
pgr2=pg/length(bl);  
subplot(246);bar(pgr2);title('中值滤波去椒盐噪声后的直方图');

r=0:255;  
bh=image_saltpepper_noise_after2(:); 
pu=hist(bh,r);  
pgr3=pu/length(bh);  
subplot(247);bar(pgr3);title('高斯平滑去椒盐噪声后的直方图');

r=0:255;  
ba=image_saltpepper_noise_after3(:); 
pa=hist(ba,r);  
pgr4=pa/length(ba);  
subplot(248);bar(pgr4);title('维纳滤波去除椒盐噪声后的直方图');
~~~

# 四，高斯噪声去除能力对比

### 1，去噪效果对比

![](https://docs.gechiui.com/gc-content/uploads/sites/kancloud/2016-03-02_56d6526aa186a.jpg)
![](https://docs.gechiui.com/gc-content/uploads/sites/kancloud/2016-03-02_56d6526ac1bee.jpg)
![](https://docs.gechiui.com/gc-content/uploads/sites/kancloud/2016-03-02_56d6526ae8bee.jpg)

### 2，实验代码

~~~
%读入原始图像并显示
image_original=imread('dog.bmp');
figure(1)
subplot(2,4,1);
imshow(image_original);
title('原输入图像');
axis square;

%生成含高斯噪声图像并显示
pp=0.05;
image_gaosi_noise=imnoise(image_original,'gaussian',0,pp);

%生成含椒盐噪声图像并显示
%dd=0.01;
%image_saltpepper_noise=imnoise(image_original,'salt & pepper',dd);

%生成含乘性噪声图像并显示
%var=0.05;
%image_speckle_noise=imnoise(image_original,'speckle',var);

image_gaosi_noise_after1=medfilt2(image_gaosi_noise,[3,3]);
subplot(2,4,2);
imshow(image_gaosi_noise_after1);title('中值滤波去高斯噪声效果图');
axis square;

h_gaosi1=fspecial('gaussian',3,1);
image_gaosi_noise_after2=imfilter(image_gaosi_noise,h_gaosi1);
subplot(2,4,3);
imshow(image_gaosi_noise_after2);title('高斯平滑去高斯噪声效果');
axis square;

image_gaosi_noise_after3=wiener2(image_gaosi_noise,[5 5]);
subplot(2,4,4);
imshow(image_gaosi_noise_after3);title('维纳滤波去高斯噪声效果');
axis square;

%原图像直方图
r=0:255;  
bb=image_original(:); 
pg=hist(bb,r);  
pgr1=pg/length(bb);  
subplot(245);bar(pgr1);title('源输入图像的直方图');

r=0:255;  
bl=image_gaosi_noise_after1(:); 
pg=hist(bl,r);  
pgr2=pg/length(bl);  
subplot(246);bar(pgr2);title('中值滤波去高斯噪声后的直方图');

r=0:255;  
bh=image_gaosi_noise_after2(:); 
pu=hist(bh,r);  
pgr3=pu/length(bh);  
subplot(247);bar(pgr3);title('高斯平滑去高斯噪声后的直方图');

r=0:255;  
ba=image_gaosi_noise_after3(:); 
pa=hist(ba,r);  
pgr4=pa/length(ba);  
subplot(248);bar(pgr4);title('维纳滤波去除高斯噪声后的直方图');
~~~

# 五，乘性噪声去除能力对比

### 1，去噪效果对比

![](https://docs.gechiui.com/gc-content/uploads/sites/kancloud/2016-03-02_56d6526b0fcaf.jpg)
![](https://docs.gechiui.com/gc-content/uploads/sites/kancloud/2016-03-02_56d6526b2a4a6.jpg)
![](https://docs.gechiui.com/gc-content/uploads/sites/kancloud/2016-03-02_56d6526b3f248.jpg)

### 2，实验代码

~~~
%读入原始图像并显示
image_original=imread('dog.bmp');
figure(1)
subplot(2,4,1);
imshow(image_original);
title('原输入图像');
axis square;

%生成含高斯噪声图像并显示
%pp=0.01;
%image_gaosi_noise=imnoise(image_original,'gaussian',0,pp);

%生成含椒盐噪声图像并显示
%dd=0.01;
%image_saltpepper_noise=imnoise(image_original,'salt & pepper',dd);

%生成含乘性噪声图像并显示
var=0.01;
image_speckle_noise=imnoise(image_original,'speckle',var);

image_speckle_noise_after1=medfilt2(image_speckle_noise,[3,3]);
subplot(2,4,2);
imshow(image_speckle_noise_after1);title('中值滤波去乘性噪声效果图');
axis square;

h_gaosi1=fspecial('gaussian',3,1);
image_speckle_noise_after2=imfilter(image_speckle_noise,h_gaosi1);
subplot(2,4,3);
imshow(image_speckle_noise_after2);title('高斯平滑去乘性噪声效果');
axis square;

image_speckle_noise_after3=wiener2(image_speckle_noise,[5 5]);
subplot(2,4,4);
imshow(image_speckle_noise_after3);title('维纳滤波去乘性噪声效果');
axis square;

%原图像直方图
r=0:255;  
bb=image_original(:); 
pg=hist(bb,r);  
pgr1=pg/length(bb);  
subplot(245);bar(pgr1);title('源输入图像的直方图');

r=0:255;  
bl=image_speckle_noise_after1(:); 
pg=hist(bl,r);  
pgr2=pg/length(bl);  
subplot(246);bar(pgr2);title('中值滤波去乘性噪声后的直方图');

r=0:255;  
bh=image_speckle_noise_after2(:); 
pu=hist(bh,r);  
pgr3=pu/length(bh);  
subplot(247);bar(pgr3);title('高斯平滑去乘性噪声后的直方图');

r=0:255;  
ba=image_speckle_noise_after3(:); 
pa=hist(ba,r);  
pgr4=pa/length(ba);  
subplot(248);bar(pgr4);title('维纳滤波去除乘性噪声后的直方图');
~~~
  

# 六，PNSR客观对比

（PNSR客观对比越高越好）

本对比也囊括了其他常见去噪方式的对比

![](https://docs.gechiui.com/gc-content/uploads/sites/kancloud/2016-03-02_56d6526b5f9bd.jpg)
  

  

  

  

#参考资源

【1】《百度百科》

【2】《维基百科》

【3】冈萨雷斯《数字图像处理》

【4】http://blog.csdn.net/ebowtang/article/details/38960271